Criterio 9. Resultados claves de la organización
4. La Medicina Intensiva y la Calidad: un reto hecho realidad
A good photograph almost always has a full range of tones, from black (or nearly black) to white (or nearly white). See Chapter 5, “ Restoring Tone, ” for more elaboration on this point, but for the time being take it as gospel.
Correspondingly , a good scan of a deteriorated photograph spans most of the range of values from near-black to near-white. It doesn’t throw away any of the intermediate tones in the photograph by forcing them to pure white or pure black. This is true for color as well as B & W photographs. A good scan’s histogram looks like the middle one in Figure 4-5 ; you don’t want it to look like the top or bottom histograms. The former makes poor use of the range of available values; the latter clips some of the near-whites and near-blacks.
It ’s usually a bad idea to make a scan that faithfully reproduces a deteriorated photograph. The upper photo in Figure 4-6 is a straight, uncorrected 8-bit scan from
Fig. 4-5 A good scan uses most of the range of available tonal values. The histogram at the top is from a scan that is too low in contrast; barely half the full range of 256 values is being used. The middle histogram shows an ideal scan; most of the values have some data, but no information is being clipped off. The histogram at the bottom shows a scan that is too contrasty. There are big spikes at values of 0 and 255, which means that some light and dark tones have been forced to pure B & W. That highlight and shadow information is lost forever; avoid that in your scans.
Fig. 4-6 A good scan is vitally important to doing a good restoration. The upper photo is a straight 8-bit scan from a badly faded original, similar in appearance to the original photograph. It is a very poor place to start from because it doesn’t take advantage of the full range of tonal values that are available in the scan. I adjusted the scanner software’s Curves and Levels controls in each color channel in the scanner software to produce a good range of tones for all three colors. (I desaturated the image in Photoshop to eliminate a small amount of lingering color cast.) That produced the good 8-bit scan in the bottom photograph; it has a histogram like the middle one in Figure 4-5 .
a faded B & W print. This is not a good basis for a restoration. The upper histogram in Figure 4-7 shows why. Only half the total range of available values is actually being used in this scan. I can expand the tonal range of the scan in Photoshop to restore the photograph to a full-valued, neutral-colored image, but if I do that I get one of those unwelcome “ picket fence ” histograms ( Figure 4-7 , bottom).
There are many gaps in the tonal scale that will show up as discontinuous- tone steps in the print ( Figure 4-8 , left). The bottom photo in Figure 4-6 is from
Fig. 4-7 These histograms show the problems that a poor 8-bit scan will cause. The top histogram is from the top scan in Figure 4-6 . The bottom histogram shows what happens when I attempt to restore that scan in Photoshop to produce a photograph with a full range of tones from black to white. There are many gaps in the histogram because there aren’t enough distinct gray levels to fi ll them in. This “ picket fence ” problem degrades the quality of the restoration.
Fig. 4-8 Enlarged portions of fi nished restorations from the straight (left) and optimized scans (right) in Figure 4-6 . The “ picket fence ” effect degrades the tonal quality of the left restoration. Only about half the normal number of gray levels are available in the scan. The skin tones look sandy instead of smooth because there are intermediates tones missing. The restoration from the good scan shows very smooth tonality because it uses almost all the available tonal levels.
a good scan, one where I adjusted the Curves and Levels controls in the scanner software to produce an image that had a much more complete and neutral range of tones. When I use that as the basis for my restoration, I get results like those in Figure 4-8 , right; here there is good, continuous tonal quality.
Don ’t be obsessed by the histogram. Some folks feel that if there are any gaps at all in their photograph’s histogram, the quality of the output will be terribly compromised. That’s an extreme exaggeration; a moderate number of gaps are almost never visible in the print. The examples I’m presenting here are extreme so that I can make it clear that many gaps are not a good thing, but you can have a pretty ratty-looking histogram and still see excellent tonality in the fi nal print. Unless I’ll be expanding or compressing the tonal scale by more than 25% when I work on the fi le, I don’t worry about gaps.
How to scan a faded B & W print
Any time you use your image processing software to expand the range of tones that you have in the scan, you’re going to get some gaps in the gray scale. That’s normal, and a few gaps in the histograms really won’t be visible in a print. What you want to avoid, though, are lots of gaps in the histogram, as in the lower portion of Figure 4-7 , or a histogram that doesn’t make good use of the entire value range, as shown on the left in Chapter 5, Figure 5-2. The more data you have to work with in your restoration, the better off you’ll be.
There are two routes to better, fuller histograms. The fi rst is to do all your scanning in 16-bit mode; that’s my habit. Even when the original has a very narrow tonal range, as in Figure 4-6 , a 16-bit scan will usually capture enough gray levels to produce a fully populated histogram when the tonal scale is expanded to produce a normal range of densities from black to white. That produces much better tonality in the fi nished restoration.
The second solution is to adjust the Levels and Curves controls in the scanner software to produce a good range of data in the scan. Make adjustments to each of the individual RGB channel’s Levels controls in the scanner software, as in Figure 4-21 . Your goal is to use the majority of the value range in each RGB channel.
Best of all, though, is to use both approaches: Scan in 16-bit mode and adjust levels to optimize your scans. You’ll be amazed how much easier it makes the restoration process.
By the way, it’s true that most scanners (and digital cameras, for that matter) don’t produce true 16-bit fi les. That is, they don’t actually produce 2 16 (65,000) distinct gray levels for each color channel. More commonly they produce 10 or 12 bits ’ worth of clean and distinct tonal information. The really good ones may produce 14. But that’s plenty! Even 10 bits of clean data is 1000 gray levels — there will be four gray levels in that fi le for every one that you’d have in an 8-bit scan. That means you could expand portions of the tonal range by a
factor of four before you would start to see gaps in the histogram. I expanded the range of tones in Figure 4-7 by only half that much (a factor of two), and that was a major adjustment. Expanding the tonal scale by a factor of four is enough to correct the most seriously faded photograph. So, even a few extra bits of data in each pixel are suffi cient to give you full histograms with few gaps, no matter how much you manipulate the photograph.
All scanners collect data internally to more than 8 bits, even if you’ve set the controls to output an 8-bit scan. The scanner has plenty of extra value levels available to fi ll in any holes, just as you do when you work on a 16-bit fi le in the computer. You can greatly expand or compress parts of the tonal range in the scanner settings without producing tonal gaps in an 8-bit output. This is why you should try to get the levels approximately right in the scam, especially if you’re going to be working with 8-bit fi les.
How to scan a dark B & W print
Here ’s how making the scanner adjustment works in practice: The B & W print in Figure 4-9 has a very limited range of tones because of fading and staining. A straight 8-bit scan with no special corrections produces the upper histogram in Figure 4-10 . You can see that this doesn’t come close to taking advantage of the full number of levels available in Photoshop.
Figure 4-11 shows the Levels adjustments I made in my scanner software. Observe how I’ve pulled in the white and black sliders so that they more closely bracket the range of tones in the print. Allow yourself some safety margin. Keep the darkest pixels in the scan at values of 10 to 20 and the lightest pixels around 240. That way you’ll avoid accidentally clipping the highlights or shadows. If you need a pure white or a pure black in the fi nished restoration, you can adjust the range in your computer without visibly compromising tonal quality.
The results of this adjusted scan are clearly much better both visually and data-wise ( Figure 4-12 ). The histogram looks much fuller ( Figure 4-10 , lower). This would be a good starting point for restoration in either an 8-bit or a 16-bit scan.
Even when I’m working with a “ B & W ” original, I still scan in full-color mode. A scanner’s B & W mode usually uses only one channel of data (green). That yields noisier scans than scanning in full-color mode and combining the channels in your image processing program. It’s easy enough to convert the image to monochrome once it’s in your computer, and the quality will be much better.
Don ’t worry about the exact image color when you’re scanning monochrome originals. Once you’ve got the scan in the computer, you’ll desaturate it
Fig. 4-9 This photograph is very dark and low in contrast. In all probability it was printed poorly to begin with, but that’s fi xable with digital restoration.
Fig. 4-10 The top histogram shows how little tonal information there is in Figure 4-9 . Hardly even one-third of the histogram bins contain any data. The bottom histogram represents Figure 4-11 , which was made using the scanner Levels settings shown in Figure 4-12 .
adjust the color to replicate the look of an original, pristine print. I tell you several ways to do that in Chapter 12, “ Printing Tips. ”
Color is a valuable tool for restoring B & W photographs. Differences in color in different parts of the photograph are evidence of damage or deterioration. Color is a distinguishing factor you can use to create masks that select especially damaged areas for restorative work (see Chapter 7, “ Making Masks, ” page 226) . You can also use that differential color information to fi x the damaged parts of the photograph (Chapter 8, “ Damage Control, ” page 266) . There are even advantages to scanning monochrome photographs with exaggerated and unrealistic color. Exaggerated hues in the photograph can be most useful for creating masks that isolate areas of tarnish, tape marks, and stains ( Figure 4-13 ). Once the areas are selectively masked, you can apply corrections to those areas, separately from the rest of the image. This is an extremely effective way of eliminating surface tarnish and silvering-out. (You can read about a full restoration of the photo from Figure 4-13 in Chapter 8, starting on page 280 .)
Here ’s another reason for not scanning your prints in B & W mode: The color channel the scanner uses for B & W scanning is not necessarily the best one to use for a restoration. That’s a choice you, not an unthinking piece of hardware, need to make. In Figure 4-14 the red channel is by far the most useful for performing a clean restoration. The damage to the original photograph is substantially less visible in this channel because the stains are primarily orange-yellow in color.
Fig. 4-11 These scanner Levels settings turned Figure 4-9 into Figure 4- 12 . Observe how the B & W set points are positioned just outside the range of populated bins in the histogram. That insures that no shadow or highlight detail gets clipped.
Fig. 4-12 This is a much better scan of the photograph in Figure 4-9 , looking more normal and attractive (and revealing a great deal of physical damage in the photograph).
Fig. 4-13 The fi gure on the left is the original, tarnished photograph. I exaggerated the color saturation in the scan I made (middle) so that the tarnish would stand out clearly. I used that exaggerated color difference to make a mask that selected for the tarnish, making it much easier to correct the damage and restore the photograph (right).
Fig. 4-14 Most B & W photographs will show less damage in one color channel than in the others. This photograph is very badly stained and faded (left). Most of the stains, though, are orange-yellow in color, which means that they hardly show up at all in the red channel of the scan (right). That’s the channel to use as the basis for a good B & W restoration.
(By the way, there’s more to this photo than meets the eye here. The original was so large that I had to scan it in three parts and stitch it together. See Chapter 9, “ Tips, Tricks, and Enhancements, ” for instructions on how to do that, and
Chapter 11, “ Examples, ” for a complete description of how I restored this photograph.)
When I’m ready to convert the scan of a monochrome photograph to grayscale, I almost never use the Desaturate adjustment or grayscale mode conversion. Instead, I use the Channel Mixer with the monochrome option selected.
Why ? It’s about control. Desaturate simply mixes equal amounts of all three color channels. It’s a good choice when all three channels in your scan are of equal quality, with none signifi cantly noisier or showing more dirt and stains than the others. Torn or cracked photographs that otherwise are clean and unfaded may have these qualities, but badly deteriorated B & W photographs usually don’t.
Automatic grayscale conversion does a weighted mix of the three channels, about 60% green, 30% red, and 10% blue. This will hardly ever be the
optimum mix for converting a scan. Channel Mixer, however, lets me specify how much of each color channel goes into the grayscale version of the photograph. Sometimes one channel is obviously superior to the others, such as in Figure 4-14 , in which case I might use 100% of that channel. At other times one channel may be signifi cantly worse than the other two, so I’ll use none of that channel and some weighted mix of the other two (it doesn’t have to be 50/50%).
On occasion I’ll even use a negative percentage in one or more of the channels! That’s allowed, and useful ( Figure 4-15 ), so long as all three channels still add up to about 100%. As I explain on page 267 in Chapter 8, “ Damage Control, ” applying a negative value in the channel that most strongly shows the damage to the original photograph can markedly suppress that damage by subtracting it out of the converted image.
How to convert an RGB scan to grayscale
In Photoshop, click Image/Adjustments/Channel Mixer.... That opens up a control panel that displays the conversion options ( Figure 4-16 ). Click the Monochrome option at the bottom of the panel. Although the fi le stays in RGB mode, it turns gray. The controls default to a mixture of 40% red, 40% green, and 20% blue channels to determine the gray value of each pixel, but it’s not likely you’ll want to use that for restoration work.
Adjust the channel sliders to emphasize the color(s) you want to most retain and deemphasize the one(s) you want to get rid of. Usually you’ll want the sum of all three channels to add up to 100%, and negative values are allowed. Pay attention to the histogram display to make sure you’re not accidentally clipping shadows or highlights. When you’re happy with the results, click OK, and the image will be converted to gray. (It’s still an RGB fi le; it’s just that all the channels have the same values.)
How you do the conversion to monochrome (assuming that’s what you want to do) depends on the characteristics of the photograph and exactly how you plan to restore it. A conversion that eliminates one or more color channels of data because they emphasize defects is a good place to start a restoration. Conversely, a conversion that emphasizes the channels that show damage most
Fig. 4-15 The original photograph (upper) is faded and stained. Examining the individual color channels shows that the red channel (middle) is the most free of damage, while the blue channel shows it the worst. The bottom photograph shows what happens when I use Channel Mixer to subtract the dirty blue channel from the cleaner red channel. The result has clearer subject detail and is cleaner than the pure red channel.
clearly will be useful for constructing selections and masks to isolate the damaged areas for repair (Chapter 7, page 230) .
For these reasons, don’t gratuitously throw away information. Even if you’re sure that the fi nal photograph will be reconstructed from only one color channel, save the fi le with all three channels. You might fi nd that the other two channels contain information that’s going to help you in other ways.